|
Résumé :
|
Despite the abundance of approaches based on the Electroencephalogram (EEG) signal, many fails to achieve real-world applicability on account of insufficient generalization capability. Moreover, recent studies have begun to adopt Deep Learning methods, which despite attaining remarkable performance when compared to traditional classifiers, result in a loss of clinical interpretability, hindering their implementation on medical devices. This work concerns the development of a prediction scheme based on pre-surgical monitoring data from the CHB-MIT database, employing three different signal processing methods for Convolution Neural Network model prediction. Namely, we experimented with Short Time Fourier Transform, Continuous Wavelet Transform and Synchro squeezed Wavelet Transform. Considering a group of 24 patients suffering from Epilepsy, the proposed methodology was performed on only 5 of them, loss, accuracy, recall and precision scores were also obtained. The results demonstrate the possibility of identifying the pre-ictal period.
|
Accueil
